This repository hosts the Epoch Based Memory Reclamation System for the Chapel programming language. We present 1) AtomicObjects, a user-facing abstraction for performing atomics on arbitrary unmanaged objects, including objects which happen to be remote; 2) LocalEpochManager and EpochManager, a scalable shared-memory and distributed-memory implementation of epoch-based reclamation that enable concurrent-safe memory reclamation of objects arbitrary unmanaged objects; 3) LockFreeQueue, a shared-memory implementation of Michael and Scott's Queue using the LocalEpochManager which demonstrates its usage and acts as its own stand-alone data structure contribution; 4) LockFreeStack, a shared-memory implementation of Treiber Stack using the LocalEpochManager.
Documentation can be seen here.
This project was made possible through the Google Summer of Code program, which allowed me the opportunity to develop new solutions in the area of distributed computing (PGAS in particular). I would also like to thank my mentors, Louis Jenkins (@LouisJenkinsCS) and Michael Ferguson (@mppf). Finally, I would like to thank the Chapel project and Cray for providing me this learning opportunity and providing me access to Cray-XC50 cluster, which allowed us to benchmark the project.
Discussions can be seen here. Pull Request can be found here.
I would like to mention a bug we discovered in chapel while working on the project, which is a bug in Forward Cycle Detection and Serial Loops. When forwarding, it seems that the compiler will detect a cycle when there is none. This only occurs in a serial loop (hence for not forall); as well, even more strangely, is that it only appears if you invoke two methods (or more) on the object being forwarded to. The bug however was non-critical for the project. Link.
The AtomicObjects module is a tested package that provides the ability to perform atomic operations on arbitrary unmanaged objects. To circumvent the infamous ABA problem, we provide an ABA wrapper that contains both the object as well as an ABA counter; all atomic operations on an ABA counter is performed via Intel's CMPXCHG16B assembly instruction. This is used to implement the LocalEpochManager and the distributed EpochManager.
Epoch-Based Memory Reclamation, when decoupled with Read-Copy-Update, is a very powerful and the current state-of-the-art in research literature data structures, and is comparable to other styles of memory reclamation. Epoch-based memory reclamation offers the ability to both denote lifetimes via 'epochs', in which unlike hazard pointers or shared atomic reference counting, it is not restricted to controlling the lifetime of a single object at a time.
An epoch manager is intended to be restricted to a specific data structure, for example a non-blocking data structure. We offer both a shared-memory optimized variant, LocalEpochManager, as well as a global-view distributed variant EpochManager that makes use of privatization.
LockFreeQueue is an implementation of Michael & Scott's Queue in Chapel that makes use of the LocalEpochManager internally. To show the performance benefits of the EpochManaged, we present a small table that shows performance comparisons to the the "Epoch Managed MS-Queue", what is being presented here as a contribution; to the 'Two-Lock' variant of the Michael & Scott Queue that uses Test-And-Set loop and one which uses a Test-And-Test-And-Set loop, both being variants of spinlocks, and finally one recycling memory and using the ABA wrapper introduced in AtomicObjects module.
| Test | Time |
|---|---|
| Epoch Managed MS-Queue | 37.7992s |
| Two-Lock (TATAS) MS-Queue | 69.8408s |
| Two-Lock (TAS) MS-Queue | 148.661s |
| Recycled (ABA) MS-Queue | 154.375s |
The module will automatically try to reclaim after every so often, but also exposes the ability to explicit trigger a 'garbage collection' and forward the epoch.